Scholarlink

A degenerative and geriatric disease like Alzheimer’s remains a conundrum despite many a novel advance in the field of medicine and physiology. The proposed work endeavors to bring out super-resolution images from the magnetic resonance imaging (MRI) scans to enable the medical fraternity to use these improvised images with augmented clarity. The graphical neural network (GNN) model attempts to enhance the lower-resolution images through super resolution and train the images of Alzheimer’s disease (AD) as gleaned from Kaggle to test its efficacy in predicting the disease through analysis of the input images. The model is found to perform well as the performance metrics, namely, the PSNR values of 30.279, 32.875, and 34.749; SSIM values of 0.8142, 0.8935, and 0.9167; and classification accuracy values of 99.51%, 98.94%, and 98.63% using super-resolution factors 2, 4, and 6.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Graphical Neural Network (GNN) Model for Super-Resolution Images to Detect Alzheimer’s Disease

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Pujari Venkata Yeswanth,
  • Samudrala Hareesh,
  • Y. Thanya,
  • K. M. Lokesh Kumar,
  • Benraji K. V. Gopal Morampudi,
  • S. Deivalakshmi

摘要

A degenerative and geriatric disease like Alzheimer’s remains a conundrum despite many a novel advance in the field of medicine and physiology. The proposed work endeavors to bring out super-resolution images from the magnetic resonance imaging (MRI) scans to enable the medical fraternity to use these improvised images with augmented clarity. The graphical neural network (GNN) model attempts to enhance the lower-resolution images through super resolution and train the images of Alzheimer’s disease (AD) as gleaned from Kaggle to test its efficacy in predicting the disease through analysis of the input images. The model is found to perform well as the performance metrics, namely, the PSNR values of 30.279, 32.875, and 34.749; SSIM values of 0.8142, 0.8935, and 0.9167; and classification accuracy values of 99.51%, 98.94%, and 98.63% using super-resolution factors 2, 4, and 6.